Before this change we were setting the headers on the PUT
request for normal and multipart uploads. For normal uploads this caused the error
403 Forbidden: There were headers present in the request which were not signed
After this fix we set the headers in the object upload request itself
as the s3 SDK expects.
This means that we only support a limited range of headers
- Cache-Control
- Content-Disposition
- Content-Encoding
- Content-Language
- Content-Type
- X-Amz-Tagging
- X-Amz-Meta-
Note for the last of those are for setting custom metadata in the form
"X-Amz-Meta-Key: value".
This now works for multipart uploads and single part uploads
See also #59
* s3: add `max_upload_parts` support
This allows to configure a maximum amount of chunks used to upload file:
- Support Scaleway which has a limit of 1k chunks currently
- Reduce a cost on S3 when each request costs some money at the expense of memory used
Co-authored-by: Nick Craig-Wood <nick@craig-wood.com>
Before this change, attempting to upload a single file into an s3
bucket which did not have create permission gave AccessDenied: Access
Denied error when it tried to create the bucket.
This was masked until e2bf91452a was
fixed.
This fix marks the bucket as OK if a fetch on an object indicates it
is OK. This stops rclone thinking it has to create the bucket in the
first place.
Fixes#4297
Previously we had a map of pools for different chunk sizes.
In practice the mapping is not very useful and requires a lock.
Pools of size other that ChunkSize can only happen when we have a huge file (over 10k * ChunkSize).
We need to have a bunch of identically sized huge files.
In such case most likely ChunkSize should be increased.
The mapping and its lock is replaced with a single initialised pool for ChunkSize, in other cases pool is allocated and freed on per file basis.
In 5470d34740 "backend/s3: use low-level-retries as the number
of SDK retries" we switched over to using the AWS SDK low level
retries instead of rclone's low level retry logic.
This had the unfortunate attempt that retrying listings to correct XML
Syntax errors failed on non S3 backends such as CEPH. The AWS SDK was
also retrying the XML Syntax error request which doesn't make sense.
This change turns off the AWS SDK retries in favour of just using
rclone's retry logic.
Amazon S3 is built to handle different kinds of workloads.
In rare cases where S3 is not able to scale for whatever reason users
will face status 500 errors.
Main mechanism for handling these errors are retries.
Amount of needed retries varies for each different use case.
This change is making retries for s3 backend configurable by using
--low-level-retries option.
Currently each multipart upload allocated his own buffers, which after
file upload was garbaged. Next files couldn't leverage already allocated
memory which resulted in inefficent memory management. This change
introduces backend memory pool keeping memory chunks which can be
used during object operations.
Fixes#3967
The error code 500 Internal Error indicates that Amazon S3 is unable to handle the request at that time. The error code 503 Slow Down typically indicates that the requests to the S3 bucket are very high, exceeding the request rates described in Request Rate and Performance Guidelines.
Because Amazon S3 is a distributed service, a very small percentage of 5xx errors are expected during normal use of the service. All requests that return 5xx errors from Amazon S3 can and should be retried, so we recommend that applications making requests to Amazon S3 have a fault-tolerance mechanism to recover from these errors.
https://aws.amazon.com/premiumsupport/knowledge-center/http-5xx-errors-s3/
The S3 ListObject API returns paginated bucket listings, with
"MaxKeys" items for each GET call.
The default value is 1000 entries, but for buckets with millions of
objects it might make sense to request more elements per request, if
the backend supports it. This commit adds a "list_chunk" option for
the user to specify a lower or higher value.
This commit does not add safe guards around this value - if a user
decides to request a too large list, it might result in connection
timeouts (on the server or client).
In AWS S3, there is a fixed limit of 1000, some other services might
have one too. In Ceph, this can be configured in RadosGW.
Before this patch we were failing to URL decode the NextMarker when
url encoding was used for the listing.
The result of this was duplicated listings entries for directories
with >1000 entries where the NextMarker was a file containing a space.
Before this change we used the same (relatively low limits) for server
side copy as we did for multipart uploads. It doesn't make sense to
use the same limits since no data is being downloaded or uploaded for
a server side copy.
This change introduces a new parameter --s3-copy-cutoff to control
when the switch from single to multipart server size copy happens and
defaults it to the maximum 5GB.
This makes server side copies much more efficient.
It also fixes the erroneous error when trying to set the modification
time of a file bigger than 5GB.
See #3778
Before this change multipart copies were giving the error
Range specified is not valid for source object of size
This was due to an off by one error in the range source introduced in
7b1274e29a "s3: support for multipart copy"
Before this change rclone would allow the user to stream (eg with
rclone mount, rclone rcat or uploading google photos or docs) 5TB
files. This meant that rclone allocated 4 * 525 MB buffers per
transfer which is way too much memory by default.
This change makes rclone use the configured chunk size for streamed
uploads. This is 5MB by default which means that rclone can stream
upload files up to 48GB by default staying below the 10,000 chunks
limit.
This can be increased with --s3-chunk-size if necessary.
If rclone detects that a file is being streamed to s3 it will make a
single NOTICE level log stating the limitation.
This fixes the enormous memory usage.
Fixes#3568
See: https://forum.rclone.org/t/how-much-memory-does-rclone-need/12743
This works around a bug in Ceph which doesn't encode CommonPrefixes
when using URL encoded directory listings.
See: https://tracker.ceph.com/issues/41870
When used with v2_auth = true, PresignRequest doesn't return
signed headers, so remote dest authentication would be fail.
This commit copying back HTTPRequest.Header to headers.
Tested with RiakCS v2.1.0.
Signed-off-by: Anthony Rusdi <33247310+antrusd@users.noreply.github.com>
- Read the storage class for each object
- Implement SetTier/GetTier
- Check the storage class on the **object** before using SetModTime
This updates the fix in 1a2fb52 so that SetModTime works when you are
using objects which have been migrated to GLACIER but you aren't using
GLACIER as a storage class.
Fixes#3522
- change the interface of listBuckets() removing dir parameter and adding context
- add makeBucket() and use in place of Mkdir("")
- this fixes some corner cases in Copy/Update
- mark all the listed buckets OK in ListR
Thanks to @yparitcher for the review.
- Change rclone/fs interfaces to accept context.Context
- Update interface implementations to use context.Context
- Change top level usage to propagate context to lover level functions
Context propagation is needed for stopping transfers and passing other
request-scoped values.
Before this change rclone would fail with
Failed to set modification time: InvalidObjectState: Operation is not valid for the source object's storage class
when attempting to set the modification time of an object in GLACIER.
After this change rclone will re-upload the object as part of a sync if it needs to change the modification time.
See: https://forum.rclone.org/t/suspected-bug-in-s3-or-compatible-sync-logic-to-glacier/10187
Make the pacer package more flexible by extracting the pace calculation
functions into a separate interface. This also allows to move features
that require the fs package like logging and custom errors into the fs
package.
Also add a RetryAfterError sentinel error that can be used to signal a
desired retry time to the Calculator.
Before this change buckets were created with the same ACL as objects.
After this change, the user can set just --s3-acl to set the ACL of
buckets and objects, or use --s3-bucket-acl as well to have a
different ACL used for bucket creation.
This also logs at INFO level the creation and deletion of buckets.
The existing s3 backend passed all integration tests with OSS provided
`force_path_style = false`.
This makes sure that is so and adds documentation and configuration
for OSS.
Thanks to @luolibin for their work on the OSS backend which we ended
up not needing.
Fixes#1641Fixes#1237
Increasing the --s3-upload-concurrency to 4 (from 2) gives an
additional 45% throughput at the cost of 10MB extra memory per transfer.
After testing the upload perfoc
Before this change rclone would use multipart uploads for any size of
file. However multipart uploads are less efficient for smaller files
and don't have MD5 checksums so it is advantageous to use single part
uploads if possible.
This implements single part uploads for all files smaller than the
upload_cutoff size. Streamed files must be uploaded as multipart
files though.
Wasabi has two location, US East and US West, with different endpoint URLs.
When configuring S3 to use Wasabi, provide the endpoint information for both
locations.
When the env_auth option is enabled, the AWS SDK's session constructor
now loads configuration from ~/.aws/config and environment variables,
and credentials per the selected (or default) AWS_PROFILE's settings.
This is accomplished by **NOT** including any Credential provider in the
aws.Config passed to the session constructor: If the Config.Credentials
is non-nil, that will always be used and the user's configuration re
role_arn, credential_source, source_profile, etc... from the shared
config will be completely ignored.
(The conditional creation and configuration of the stscreds Credential
provider is complicated enough that it is not worth re-creating that
logic.)
Before this change the ACL for objects which were server side copied
was left at the default "private" settings. S3 doesn't copy the ACL
from the source when you copy an object, you have to set it afresh
which is what this does.
This unifies the 3 methods of reading config
* command line
* environment variable
* config file
And allows them all to be configured in all places. This is done by
making the []fs.Option in the backend registration be the master
source of what the backend options are.
The backend changes are:
* Use the new configmap.Mapper parameter
* Use configstruct to parse it into an Options struct
* Add all config to []fs.Option including defaults and help
* Remove all uses of pflag
* Remove all uses of config.FileGet
These are AWS, Ceph, Dreamhost, IBM COS S3, Minio, Wasabi and Other.
This configures endpoints where known and makes sure config doesn't
appear where it isn't valid where possible.
This introduces a method of making provider specific configuration
within a remote. This is useful particularly in s3.
This commit does the basic configuration in S3 for IBM COS.
From testing it appears that CEPH no longer works properly with v2
auth and neither does Dreamhost, so update the docs anc configuration
to recommend v4 auth.
In a typical rclone copy to a bucket/container based remote, before
this change we were doing a list, followed by a HEAD of the bucket to
check it existed before doing the copy. The fact the list succeeded
means the bucket exists so mark it OK at that point.
Issue #1421
The purpose of this is to make it easier to maintain and eventually to
allow the rclone backends to be re-used in other projects without
having to use the rclone configuration system.
The new code layout is documented in CONTRIBUTING.
